Marketing Research

MKTG 440

Measurement and Scaling

Housekeeping

  • GA1
    • How did it go?
    • Will post feedback before next class
    • Will meet with each team to briefly discuss on Wednesday
  • GA2
    • Start working on this now!
    • Will talk more about it next class
  • Office hours: Wednesday, 2-4pm (320S)
    • Is this a good time?
  • Questions?

Agenda

  • Scale types (nominal, ordinal, interval, ratio)
  • Measures of central tendency
  • Scale quality
  • Measuring complex ideas

Measurement

“I have been struck again and again by how important measurement is to improving the human condition”

— Bill Gates

Measurement

  • How does a social media platform measure ad audience?
  • Advertising is often paid for in terms of “views”, “clicks”, or “actions”
  • How should we define a “view”?
  • What are important properties of a “view”?
    • Initial play? Rewatch? Loops? AFK? Gameable?

Measurement

Measurement is the process of assigning numbers or labels to objects, events, or people according to pre-specified rules.

These rules should…

  • establish a one-to-one relationship between the numbers/labels and the characteristics being measured
  • be standardized (reliable) so that measurements are consistent across different observers and contexts

Scale Types

Four Types of Scales

There are four primary scales used in marketing research:

  1. Nominal scale: Categories/labels
  2. Ordinal scale: Nominal + order
  3. Interval scale: Ordinal + equal intervals
  4. Ratio scale: Interval + true zero

Each scale type

  • has different properties and appropriate uses
  • has different implications for data analysis

Nominal scales

Nominal scales use numbers as labels to identify and classify objects into distinct categories.

  • Numbers serve as labels or tags
  • Numbers have no ordering or mathematical meaning
  • Numbers are used for identification only

Examples:

  • Social Security numbers
  • ZIP codes
  • Any categorical variable
    • E.g. colors (blue = 1, red = 2, green = 3, yellow = 4, purple = 5, etc.)
    • E.g. gender identity, marital status, other demographics

A good nominal scale is MECE

  • Mutually exclusive: each item fits into only one category
  • Collectively exhaustive: a label exists for all possible items
  • MECE: Mutually Exclusive, Collectively Exhaustive
ME, not CE
1 2 3 4 5
CE, not ME
1 2 3 4 5
MECE
1 2 3 4 5

Ordinal scales

Ordinal scales take a step beyond nominal scales by ranking or ordering objects based on the relative amount of a characteristic.

  • Numbers indicate relative positions
  • Examples:
    • Service quality: Poor (1), Good (2), Excellent (3)
    • Spice level: Mild (1), Medium (2), Hot (3)
    • Income brackets: <$50k (1), $50k-$100k (2), >$100k (3)
  • Difference between ranks is not meaningful
    • Hot (3) is not 3x as spicy as Mild (1)

Interval Scales

Interval scales have the properties of ordinal scales, with equal intervals between scale points.

  • Examples: temperature (Fahrenheit/Celsius), time of day, calendar dates
  • Zero point is arbitrary
    • 0°F does not mean “no temperature”
    • Celsius vs. Kelvin?
  • Ratios are meaningless
    • 80°F is not twice as hot as 40°F

Ratio scales

Ratio scales have equal intervals (all of the properties of interval scales) and a true zero point.

  • True zero: an absence of the characteristic being measured
  • Examples: weight (lb), length (in), sales ($)
  • Ratio can be computed and have meaning
    • Store A with $1000 sales has twice the sales of Store B with $500 sales
    • Store C with $0 sales has no sales at all

Summary of measurement scales

Property Nominal Ordinal Interval Ratio
Describe ✓ ✓ ✓ ✓
Preserve order ✓ ✓ ✓
Equal intervals ✓ ✓
Natural zero ✓

Moving from left to right:

  • Information content increases
  • Data gets more precise, but greater burden on participant
  • More math and stats can be used

Income survey questions

What is your primary source of income? (Nominal)

○ Employment

○ Investments

○ Inheritance

○ Other

 

What is your income bracket? (Ordinal)

○ Less than $20,000

○ $20,000 - $50,000

○ $50,001 - $100,000

○ More than $100,000

 

What is your annual income? $_________ (Ratio)

Scale Type Practice

What is the scale type?

What is the highest level of education for the head of the household?

○ Less than high school

○ Some high school

○ High school graduate

○ Some college

○ College graduate or higher

Property Yes/No
Describe
Preserves order
Equal distance
Natural zero

What is the scale type?

Please provide the zip code of your home address: ___________

Property Yes/No
Describe
Preserves order
Equal distance
Natural zero

What is the scale type?

On an average day, how many cigarettes do you smoke?

○ Less than half a pack

○ Half to one pack

○ Over one pack

Property Yes/No
Describe
Preserves order
Equal distance
Natural zero

What is the scale type?

How likely are you to attend the Arizona Homecoming football game next year?

_______% chance (0-100)

Property Yes/No
Describe
Preserves order
Equal distance
Natural zero

Find all four scales in this picture

Property Yes/No
Describe
Preserves order
Equal distance
Natural zero

Measures of Central Tendency

Scales and central tendency

A measure of central tendency describes the location of the center of a distribution of data.

Nominal Ordinal Interval Ratio
Mode Appropriate Appropriate Appropriate Appropriate
Median Inappropriate Appropriate Appropriate Appropriate
Mean Inappropriate Inappropriate Appropriate Appropriate

Example

4,000 participants were asked their age:

Age Group Responses
0-18 250
19-34 1,320
35-49 2,745
50+ 315
  • What is the modal age in the sample?
  • What is the median age in the sample?
  • What is the average age in the sample?

Scale Quality

Two quality measures

A scale’s reliability refers to its consistency, the ability to get the same result repeatedly. A scale’s validity refers to its accuracy, whether it measures what it is supposed to measure.

high validity/
low reliability
low validity/
high reliability
low validity/
low reliability
high validity/
high reliability

Scale reliability

Ways to measure reliability

  • Test-retest reliability: same sample, same scale, two different times
  • Internal consistency: same sample, highly correlated answers between two (sets of) questions measuring the same construct
    • Reliability is measured within respondents
  • Equivalent forms: two samples, high correlation between two “similar yet different” scales to measure the same construct
    • Reliability is measured between respondents

Internal consistency example

To what extent do you agree or disagree with the following statements:

Housing in Tucson is affordable.

Strongly
Disagree
(1)


(2)


(3)


(4)
Strongly
Agree
(5)

 

…a few pages later…

 

Housing in Tucson is expensive.

Strongly
Disagree
(1)


(2)


(3)


(4)
Strongly
Agree
(5)

Equivalent form example

Exam 1 — Version A
87%
avg score
50% of students
Exam 1 — Version B
88%
avg score
50% of students

Measuring validity

Validity refers to the accuracy of a scale in measuring what it is intended to measure.

  • Face validity: extent to which a scale appears to measure the intended construct, as determined by the researcher’s intuition
    • Does it look like it is measuring what it is supposed to be measuring?
    • E.g., a customer satisfaction survey should ask about satisfaction, not unrelated topics
    • Requires subjective judgment
  • Content validity: extent to which a scale covers all aspects of the construct
    • E.g., measuring customer satisfaction should include multiple dimensions (product quality, service, price, etc.)
    • Requires expert judgment to evaluate whether scale items are representative

Sensitivity

Sensitivity refers to a scale’s ability to detect differences that exist in the construct being measured.

Insensitive
Sensitive
"The price of a Subaru as compared to competitors is too high"
○ Agree   ○ Disagree
○ Strongly Agree
○ Mildly Agree
○ Neutral
○ Mildly Disagree
○ Strongly Disagree
"What is today's temperature?"
○ Hot   ○ Cold
_____ °F

Discussion: Should we always use the most sensitive scale possible?

Measuring Complex Ideas

How can we measure brand loyalty?

  • Thoughts?

  • The problem: There is no single definition or measure of brand loyalty, unlike something like age or income.

  • Possible solution: Use multiple (more easily measured) indicators that together reflect the underlying construct of brand loyalty. The more our conglomerate of factors is positive, the more confident we are that brand loyalty is also positive.

Constructs

Constructs are abstract concepts that are not directly observable or measurable

  • E.g., customer satisfaction, brand loyalty, perceived quality, trust
  • While not easy to measure, they are important metrics in marketing research

Measuring constructs

  1. Construct selection/development: What is the construct we want to measure?
    • Brand loyalty
  2. Identify dimensions/factors: What are the key components of brand loyalty?
    • Factors can be abstract (attitudes, satisfaction) or concrete (repeat purchase behavior)
Brand Loyalty Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6

Aside: Measuring attitudes with Likert scales

How satisfied are you with the quality of our products?

Very
dissatisfied
(1)
Somewhat
dissatisfied
(2)

Neutral
(3)
Somewhat
satisfied
(4)
Very
satisfied
(5)
  • What type of scale is this?

Likert scales cont.

  • Typically use 5, 7, or 9 response options
    • Why an odd number?
  • We typically treat Likert scale data as interval data for analysis purposes
    • Step 1: Assign numbers to responses (e.g., 1 to 5)
    • Step 2: Assume equal intervals between points
    • Allows us to calculate means, standard deviations, etc.
  • Can combine multiple Likert items to create a composite score for a construct

Example: Attitude toward Wikipedia

Strongly
disagree
(1)		 Disagree

(2)		 Neutral

(3)		 Agree

(4)		 Strongly
agree
(5)
1. Wikipedia has a user-friendly interface. ○ ○ ○ ⊗ ○
2. Wikipedia is usually my first resource for research. ○ ○ ○ ○ ⊗
3. Information on Wikipedia is very accurate. ○ ○ ⊗ ○ ○
4. Wikipedia is a great service for society. ○ ○ ○ ○ ⊗

Calculating a composite score

Strongly
disagree
(1)		 Disagree

(2)		 Neutral

(3)		 Agree

(4)		 Strongly
agree
(5)		 Score
1. Wikipedia has a user-friendly interface. ○ ○ ○ ⊗ ○ 4
2. Wikipedia is usually my first resource for research. ○ ○ ○ ○ ⊗ 5
3. Information on Wikipedia is very accurate. ○ ○ ⊗ ○ ○ 3
4. Wikipedia is a great service for society. ○ ○ ○ ○ ⊗ 5
17

Overall attitude score: 17/20 or 4.25/5

Commonly used Likert scales

Construct (1) (2) (3) (4) (5)
Attitude Very Bad Bad Neither Bad Nor Good Good Very Good
Importance Not at All Important Not Important Neutral Important Very Important
Satisfaction Very Dissatisfied Somewhat Dissatisfied Neither Dissatisfied Nor Satisfied Somewhat Satisfied Very Satisfied
Agreement Strongly Disagree Disagree Neither Agree Nor Disagree Agree Strongly Agree

Note: There are many other examples on Qualtrics